Hemostasis-on-a-chip / incorporating the endothelium in microfluidic models of bleeding

Abstract Currently, point-of-care assays for human platelet function and coagulation are used to assess bleeding risks and drug testing, but they lack intact endothelium, a critical component of the human vascular system. Within these assays, the assessment of bleeding risk is typically indicated by the lack of or reduced platelet function and coagulation without true evaluation of hemostasis. Hemostasis is defined as the cessation of bleeding. Additionally, animal models of hemostasis also, by definition, lack human endothelium, which may limit their clinical relevance. This review discusses the current state-of-the-art of hemostasis-on-a-chip, specifically, human cell-based microfluidic models that incorporate endothelial cells, which function as physiologically relevant in vitro models of bleeding. These assays recapitulate the entire process of vascular injury, bleeding, and hemostasis, and provide real-time, direct observation, thereby serving as research-enabling tools that enhance our understanding of hemostasis and also as novel drug discovery platforms. Plain Language Summary The human body’s response to stop bleeding after a vascular injury involves a complex but finely tuned cascade of interactions between the blood, the blood vessel wall, and the physical flow of the blood. Accordingly, in vitro models that incorporate those aspects that occur in vivo are highly needed for research and clinical purposes. Here, we review the state of the art of these technologies, hemostasis-on-a-chip devices that aim to achieve those goals. These physiologically relevant “microchips” mimic the bleeding process as well as the cessation thereof, and can be leveraged as research-enabling tools, platforms for drug discovery, and clinical testing.


Introduction
Hemostasis and thrombosis are processes that occur within the vascular system.Thrombosis is defined as the pathologic intravascular blood clot formation caused by disease states of the blood and vasculature that may lead to ischemia and end-organ damage.Hemostasis is the physiological process where a blood clot formation occurs after a vascular injury to maintain blood circulation within the vasculature, physically stopping leakage of blood from the disrupted blood vessel (i.e.bleeding).Hemodynamic forces are generated upon blood vessel disruption and the hemostatic process then occurs at the site of the intravascular opening of the wound (hemostatic plug formation) and the extravascular region where blood may have leaked into.Hemostasis is only achieved when pro-and anti-coagulant processes as well as fibrinolytic/antifibrinolytic processes are finely balanced by the activity of blood components and the vasculature such as endothelial cells, and conversely, risks of excess bleeding or blood clotting may arise when those activities become imbalanced.Endothelial cells have critical roles in hemostasis.They are known to have two opposite biological functions in the vascular system: (1) Anticoagulant or thrombosuppressive property of healthy intact endothelium maintaining undisrupted blood circulation in the body.(2) Procoagulant or thrombogenic property exhibited when they are damaged and/or the endothelium is disrupted.The later property helps stop bleeding by initiating a hemostatic plug in the process of normal hemostasis, or cause pathological intravascular thrombosis.Endothelial cells can dictate when, where, and how hemostasis occur in the body.
Bleeding disorders (i.e.hemophilia and von Willebrand disease) affect hemostasis, and affected patients may exhibit various degrees of bleeding problems.Historically, patients with a suspected bleeding disorder underwent a "bleeding time test" to directly assess the patient's bleeding risk.This test includes small incisions on the patient's arm and recording the time when the bleeding stops [1] but the procedure is difficult to control and is invasive.Currently, the bleeding time test has largely been replaced by assays using blood collected through a venipuncture, assessing the risks of abnormal bleeding and blood clotting.Specifically, the assays include prothrombin and partial thromboplastin time, platelet function assays, and various viscoelastic tests.The assays can be performed as point-of-care in clinical settings [2][3][4] where timely assessment is needed in the emergency department, surgery, trauma involving hemorrhagic shock, and more recently, COVID-19associated coagulopathy [5].They can determine the patient's hemostatic competence more precisely but other critical factors that control hemostasis in the body may be omitted.For example, the fluid dynamics of blood flow and blood interaction with the cells composing the vasculatures, i.e. endothelial cells and the underlying matrix.Additionally, the assays can determine the risk of the blood clotting through direct observation of platelet behaviors and coagulation, but bleeding risks are only indicated by the lack of or the reduced amount of activity.There is no direct observation of "bleeding" and hemostasis to assess patient's hemostatic competence.
The recent development of microfluidic devices incorporating fluid dynamics of blood flow has offered inexpensive and reproducible platforms advancing the understanding of platelet function and coagulation in more relevant biophysical environments with minimal amounts of blood from human patients and healthy donors.Many point-of-care whole blood-based microfluidic models have been suggested to evaluate the potential risk of bleeding or blood clotting, and as drug testing platforms [6].Thrombogenic substrates such as collagen, kaolin, and tissue factor (TF) promote platelet adhesion and coagulation of perfused blood in various geometry and flow regimes of the microfluidic channels [7].The transparency of the chip material with polydimethylsiloxane (PDMS) enables the visualization of platelet aggregation and fibrin clot formation or lack thereof in perfused whole blood with real-time imaging.Although those microfluidic devices would be the next generation point-of-care blood assays, it is still important to include endothelial response to vascular injury and bleeding to correctly evaluate hemostasis in the vasculature.Therefore, it is imperative to develop an in vitro microfluidic bleeding systems that integrate all critical components of hemostasis including endothelial cell injury, vessel disruption, and bleeding.
This review discusses the latest hemostasis-on-a-chip, i.e. in vitro human whole blood-based microfluidic models to assess bleeding risk, models which include localized endothelial cell injury, followed by endothelial vessel wall disruption, bleeding, hemostatic platelet plug formation (primary hemostasis), fibrin clot formation (secondary hemostasis), and cessation of bleeding.Given the limited number of hemostasis-on-a-chip models published, we will also discuss the future direction of hemostasis-on -a-chip, variable factors yet to be investigated, and expectation for complete whole blood-perfusable vasculature-on-a-chip as an alternative to traditional "bleeding time test" and animal in vivo hemostasis models.

Hemostasis-on-a-chip: vascularized vessel-on-a-chip systems with localized endothelial injury and bleeding
Historically biomedical research on hemostasis has widely included in vivo animal bleeding models, for example murine tailbleeding assay and laser and needle injury models [8,9].Intravital imaging uses transmitted light and fluorescence to visualize blood cells, coagulation factors, and other proteins which may contribute to hemostasis during the injury.Despite in vivo animal bleeding models recapitulating multiple aspects of human hemostasis, limitations remain in translating those results to humans in a physiologically and/or clinically relevant manner [10].There are inherent differences between human and mouse platelets [11,12].Seok et al. found no correlation between human and mouse gene expression profiles when they are exposed to various inflammatory stimulants [13].More refined human cell-based in vitro assays will complement current animal models and eventually reduce the number of animal experiments.
Many factors inflict endothelial damage resulting in thrombosis: inflammation, oxidative stress, and physical stimulation [14,15].In in vitro human endothelial cell-based vascularized microfluidic models, various methods have been used to induce localized endothelial injury to study thrombosis.Some examples include: hematoporphyrin (a photosensitizer) induces local reactive oxygen species by light exposure [16,17]; local heat induction [18]; physical scratching [19], and growing endothelial cells in a limited surface area [20].Here, the injury causes exposure of the thrombogenic surface of the subendothelium, triggering thrombosis by whole blood perfusion.Those models provide insights into hemostasis because the basic mechanism of platelet adhesion and coagulation are same as in thrombosis.However, they lack the fluid dynamic of blood "bleeding" and are not truly equivalent models to the bleeding time test or in vivo mouse bleeding models.
PDMS-based microfluidic devices have well-characterized geometry and flow dynamics.The first in vitro PDMS-based microfluidic vasculature model that incorporated a "bleeding" element was described by Muthard et al. in 2012 [21].They described a permeation device with a micropost scaffold region where collagen and/or TF are coated and whole blood flows through (Figure 1A).This loosely mimics vessel leakage, and the clot permeability is measured in a controlled pressure drop under flow.A similar percolation element was introduced in PDMS-based microfluidic "bleeding chip" in the latest research by Lakshmanan et al. (2020) [23].They created three collagencoated micropillars positioned 10 µm apart, and at the junction of two orthogonal channels, a hemostatic plug is formed.They suggest that the chip could be used to evaluate bleeding risk upon the application of various doses of antithrombotic drugs in COVID-19 patients with microangiopathy.Schoeman et al. developed an "H" shaped-microfluidic channel which models hemostasis in collagen/TF-rich extravascular space (Figure 1B) [22].Here, the bleeding time, thrombus formation by platelet and fibrin(ogen) accumulation in the injury channel was quantitatively measured and compared with different treatments of prothrombotic substrates and antithrombotic inhibitors, specifically antibody against factor VIII and P2Y12 antagonist, 2-MeSAMP.Collectively, these PDMS-based microfluidic devices only evaluate the contribution of blood to hemostasis.They all lack vasculature cells.Specifically, endothelial cells are not included and therefore are missing the input to bleeding and resulting hemostasis by endothelial cells, which are always present at the site of the vascular injury in vivo.
An "endothelialized" microvessel-on-a-chip capable of purely mechanical endothelium injury and following bleeding was first developed by Sakurai et al. (2018), utilizing a pneumatic valve underneath a small portion of the endothelialized microvasculature channel, that opens the microvasculature channel into a new branching "bleeding" channel [24].Intact endothelium in the vascular channel is disrupted at the opening, and underneath extracellular matrix (ECM) proteins are exposed.Whole blood is perfused and observed for hemostatic plug formation, fibrin clot formation, and eventually cessation of the bleeding (Figure 2A).This pneumatic valve model provides clear views of endothelium local damage, bleeding, platelet adhesion and accumulation, fibrin formation, and measures bleeding time [24,26].It demonstrated excess bleeding of hemophilia A patient blood, a change in clot architecture by drug treatment (eptifibatide), von Willebrand factor (vWF)-dependent hemostasis at an arterial flow rate (2500 s −1 ), the rapid phosphatidylserine exposure on damaged endothelial cells during hemostasis [24], and involvement of complement lectin pathway, especially mannan-binding lectin (MBL) and MBL-associated serine proteases −1 in hemostasis [27].Another endothelialized hemostasison-a-chip was published by Poventud-Fuentes et al. (2021), using a microneedle to penetrate the vascular and underlining hydrogel layers (Figure 2B) [25].Upon the microneedle injury, bleeding occurs through ECM-rich, especially TF-rich, hydrogel layer under endothelialized microvasculature.Due to the hydrogel deformability, this model can demonstrate the hemostatic plug formation and wound closure by the clot contraction in the hydrogel layer, providing important insights into coagulation in damaged tissue outside vasculatures [25].Both "endothelialized" models enhance our understandings of the mechanism of hemostasis by direct observation of bleeding, platelet adhesion, and coagulation in more holistic environment with all components of the vascular system in the body and enable assessment of the effects of various treatments on the process of hemostasis.
Although both models combine endothelial cells and endothelial disruption followed by bleeding, many important differences are suggested and observed: (1) The pneumatic valve model physically opens part of the vascular channel to disrupt endothelium.This causes extended damage and loss of endothelial cells and exposure of ECM proteins surrounding the area where the opening is created.Because of this, the pneumatic valve model shows more intravascular platelet adhesion and clot formation and irregular shape of hemostatic plug, and hemostasis is achieved within 10 minutes [24].The microneedle model punctuates the vascular channel and underlying hydrogel layer with a fine needle and does not cause extended endothelial damage.There is a minimal intravascular hemostatic plug formed which is smaller than one observed in their in vivo mouse model [25].(2) The pneumatic valve model does not have pre-coated TF.Nevertheless, the model achieves hemostasis with fibrin formation.This implies that there are sites of thrombin generation other than pre-coated TF, for example, on activated/damaged platelets/endothelial cells.However, the bleeding time and magnitude of fibrin clot formation may be underestimated compared to a model with pre-coated TF [24].In the microneedle model, hemostasis is achieved only when TFrich extravascular wound is filled by platelets and fibrin clots.Without TF in the hydrogel, there is no hemostatic plug formed, and the wound does not close.Platelets contract the clot in the wound only under presence of TF [25].(3) The pneumatic valve  model uses a relatively higher flow rate of 500 s −1 as default and 2500 s −1 as arterial flow.The authors observed vWF incorporated at the wound area and with inhibitory antibody against vWF, hemostasis is impaired at arterial flow rate [24].The microneedle model uses a lower venous flow rate of 100 s −1 therefore, platelet's non-vWF-dependent collagen binding and TF-dependent fibrin clot formation are major contributions to the hemostasis and the incorporation of vWF may be small [25].Overall, the pneumatic valve model exhibits procoagulant properties of locally damaged endothelial cells, can provide insights into the interaction between damaged endothelial cells and blood, and how it affects hemostasis.The microneedle model instead uses endothelium as an anticoagulant surface during perfusion of whole blood.The contribution of endothelial procoagulant property by injury seems minimum.Their unique deformable TF-rich hydrogel could provide more biophysically relevant microenvironment for evaluating blood hemostatic functions and viscoelastic properties.

Conclusions and future perspectives
The latest developments of two vascularized hemostasis-on-a-chip models will bring new perspectives of human hemostasis research and new drug discovery platforms by adding endothelial cells and bleeding time.However, many aspects of the models are yet fully investigated.The importance of organ-specific endothelial heterogeneity has been addressed in recent years [22], but very few studies have been done on hemostasis research.Hemostasis-on -a-chip models have tested only human umbilical cord vein endothelial cells [24,25] and human aorta endothelial cells [24].The current hemostasis-on-a-chip models only contain endothelial cells and exclude smooth muscle cells and fibroblasts which compose larger blood vessels in the body.Including those other cells can possibly determine intrinsic TF in the models, vessel constriction by vascular injury, and establishing more biophysical and biochemical environment for hemostasis-on-a-chip in the future.The hemostatic response to a wider scale of injury sizes and different types of injury can be also explored in the future.In preclinical drug testing trials for bleeding disorders, it is critical to carefully evaluate drug potency and efficacy to blood and to endothelial cells.In healthy hemostasis, only damaged endothelial cells at the injury site become procoagulant and enhance the process of hemostasis.The undamaged endothelium maintains its anticoagulant property to prevent unwanted blood clot formation elsewhere.The therapeutic interventions designed to help patients with bleeding disorders normally enhance blood and endothelial procoagulant activities, but additional research is needed on the impact to endothelium upstream and downstream of the vascular injury area (i.e.abnormal coagulation or endothelial dysfunction elsewhere).Additionally, it is imperative to investigate the fine tuning of the drug dosage required for individual patient and interpatient variability.
Endothelial dysfunction is caused by various factors during a patient's lifetime and may be hereditary.Different phenotypes of endothelial cells from individual patients with hereditary bleeding or thrombotic disorders such as von Willebrand disease, hemophilia, and hereditary thrombotic thrombocytopenic purpura can be integrated into the hemostasis-on-a-chips.Patient-specific blood outgrowth endothelial cells (BOECs) have already been utilized in vessel-on-a-chip technology, assessing thrombosis and inflammation in diabetic and sickle blood disease patients [28,29].In the future, personalized hemostasis-on-a-chip models can include the patient's own endothelial cells and their own blood, to evaluate the drug efficacy and bleeding/thrombosis risks recapitulating a more accurate, personalized outcome of the blood treatment.
Vascularized hemostasis-on-a-chip assays will provide all the important aspects of diseases, from hemostatic functions of blood, endothelium dysfunction by injury, blood-endothelium interaction, bleeding, and hemostasis.Finally, more complicated vascularized organ-on-a-chip models have been developed in the last decade [30].Those models include lung-on-a-chip [31], Liver-on-a-chip [32], kidney-on-a-chip [33], Blood-brain barrier-on-a-chip [34,35], and vascularized tumor-on-a-chip [30,36].In some models, vasculature can be formed by vasculogenesis and angiogenesis of endothelial cells [30,37] and those self-assembled microvascular networks have proven to be perfusable [38].However, for the organ-on-a-chip models to become platforms for drug testing and studies on bleeding and hemostasis, there is a great need for more interdisciplinary studies from biomedical sciences and tissue engineering fields to overcome the challenge of whole blood perfusion, obtaining consistency in the sizes and geometry of the microvessels and flow dynamics, and the capability to introduce local injury, bleeding, and hemostasis.Hope remains that future organ-on-a-chip assays for hemostasis will eventually replace in vivo animal injury models.

Figure 1 .
Figure 1.PDMS-based microfluidic devices, "hemostasis-on-a-chip" to model bleeding and hemostasis.(A) Vessel leakage model to assess blood clotting and clot contraction under controlled fluid dynamics and pressure drop.The figure is reproduced from original publication [21], licensed under CC-BY-NC.(B) "H" shaped-microfluidic channel modeling hemostasis in collagen/tf-rich extravascular space.The figure is reproduced from original publication [22], licensed under CC-BY-NC.

Figure 2 .
Figure 2. "Endothelialized" microfluidic devices as hemostasis-on-a-chips to model bleeding and hemostasis.(A) Pneumatic valve model recapitulates mechanical vascular injury followed by bleeding and hemostasis.The figure is reproduced from original publication [24], licensed under CC-BY.(B) Microneedle model recapitulates hemostasis and clot contraction in collagen/tf-rich extravascular space.The figure is reproduced from original publication [25], licensed under CC-BY.